Study design and patients
The subsample analyzed was selected from the Osteoarthritis Initiative (OAI) study, a multicenter, population-based longitudinal cohort study focused on risk factors and disease biomarkers associated with the onset and progression of ROA. Eligible participants of the OAI were between 45 and 79 years old at baseline and we followed participants with OA or other inflammatory arthritis, end-stage arthritis, symptomatic knee OA for four years. Our data was from the OA occurrence subcohort, in which participants had characteristics that placed them at increased risk for developing symptomatic knee OA.
Demographic information (age, gender, ethnicity, height and weight) was already recorded at the first visit. BMI (weight/ height2 , kg/m2 ) was calculated at the first visit. Inclusion criteria were frequent knee symptoms and frequent medication for knee symptoms as well as other risk factors, weight, history of knee injury and surgery, bony enlargement of fingers, frequent knee bending, and family history of total knee replacement(TKR). Exclusion criteria were bilateral TKRs, plans to have bilateral TKR, inflammatory and rheumatoid arthritis, MRI contraindications, and participants with serious medical conditions or clinical trials that could interfere with study results. Informed consent was obtained from all participants.
Sample description
Case knees (n=178) were those that the incident ROA [Kellgren Lawrence grading (KLG)≥2] on knee radiographs during the 48-month of follow-up. This sample contained knee images at each follow-up time point. Bilateral knee of a participant could be included if both incident ROA. Case knees were matched one-to-one by gender, age (± 5 years) and radiographic status (KLG = 0 or 1 in the index knee) with 178 control knees that were free of incident ROA from baseline to 48 months. The first time occurrence of ROA was defined as P0, P1 was 1 year prior to P0, and the baseline was defined as the enrolling time.
Patients reported the history of knee injury and surgery at the enrollment visit (OAI study protocol). Knee injury was defined as history of injury that resulted in difficulty walking for at least one week,and surgery was defined as history of any knee surgery such as meniscal repairs. Knee frequently bend count was recorded during follow-up visit, using 0-5 degree points to measure the different activities of knees.
Radiographs protocol
All participants underwent fixed knee x-ray measurement at baseline and each follow-up visits. The posteroanterior projection was taken in a standing position with 20 to 30 degrees of flexion and 10 degrees of internal rotation. Reader reference standard protocols for knee radiographs including KLG. The ROA was defined as KLG ≥ 2[22]. Using 3T MRI from Philips Medical Systems, coronal T2-weighted turbo spin echo (TSE) MRI images, standardized scans were performed on the study knee joints at baseline and follow-up with echo time of 29 ms, 35 slices, section thickness of 3mm, and no gaps. MR images were assessed at P0 , P1 and BL. MRI image database was transferred to a separate workstation and measured manually by a trained reader using the mimics software version19.0 [23] and assessed by trained readers to treatment without know participants' group assignments. All MRIs were read sequentially, unblinded to the time point, but blinded to the participant clinical data and grouping status.
Magnetic resonance imaging assessment
On coronal T2-weighted turbo spin echo (TSE) MRI images, the mean value of the medial and lateral meniscal lesions distance (MLD) and tibial plateau width (TPW) on MRI slices from the tibial eminence to the posterior horn of the meniscus were measured manually and quantitatively. Mean of MLD and Mean of TPW were expressed by Mean (MLD) and Mean (TPW) respectively. The lesions distance of the medial and lateral meniscus (MLD) was defined as the distance from the inner edge of the medial meniscus lesions to the outer edge of the lateral meniscus lesions. Relative percentage of the meniscal lesions distance (RMLD) was defined as MLD divided by the width of the tibial plateau(%). [Mean(RMLD)] was the mean value of the RMLD measurement in one knee. Morphological parameters of the medial and lateral menisci were quantitatively measured at P0 , P1 and baseline, included [Mean(MLD)], [Mean(TPW)] and [Mean(RMLD)] of knee joint. Fig.1A was the normal knee joint model, showed the anatomical structure of part of the knee joint. Fig.1B was a simplified lesion model showed several types of lesions that may affect the new parameters of the meniscus. MLD was introduced to represent the distance from the most medial edge of a continuous lesion starting from the inner edge of the medial meniscus to the most lateral edge of a continuous lesion starting from the inner edge of the lateral meniscus. In other words, The white dashed line was the distance of the medial and lateral meniscal lesions (MLD). Segmentation was restricted to all slices from the tibial intercondylar spine to the posterior side of the knee, where meniscus lesions are most representative[24]. We then calculated the mean of the measured parameters at all slices from the intercondylar spine to the posterior of the knee joint (the average length of the white dashed lines in Figure 1 B).
To calculate these parameters, the measured slices in knee MRI images were identified first, and then the target values were measured layer by layer on the MRI image.
To measure MLD (Fig.2A, B), we identified point c and point d on the MRI image slice and then linked point c and point d. Point c was the most lateral edge of a continuous lesion starting from the inner edge of the lateral meniscus. Point d was the most medial edge of a continuous lesion starting from the inner edge of the medial meniscus.
To measure TPW(Fig.2B), the measurement was performed by first drawn two vertical lines (Line A and Line B ), Line B intersected the peripheral margin of the lateral tibial plateau and Line A intersected the inboard margin of the lateral tibial plateau; the distance between these two lines was defined as PTW. Any suspected osteophytes of the tibial plateau were not included in the measurement of TPW.
To calculate the Mean(RMLD), the ratio of MLD to PTW on each MRI slice was calculated separately, and then the average was taken. Respectively, Mean(MLD) and Mean(PTW) were the mean values of MLD and PTW, respectively, across all MRI measurement slices.
Statistical analysis
A trained reader measured the novel parameters in all MRI images. To calculate reliability, the novel parameters of meniscus were measured by the same reader twice with a one-month gap for intra-class reliability (intra-class correlation coefficients, ICCs) and by two readers independently for inter-class reliability (inter-class correlation coefficients)[25] in 51 randomly selected MRI images. T,and Chi-Squared tests was used to compare statistic difference between case and control groups. Conditional logistic regression analyses were used to assess the risk of incident ROA in participants before and after adjustment for BMI, previous knee surgery, previous knee injury and frequently bend count at baseline. Models run at three time points: P0, P1 and baseline. All statistical analyses were subject-based and performed using R-Studio software and P-value less than 0.05 (2- tailed) was considered as statistical significant.